Broadcom and NVIDIA to Power Private AI

Broadcom Inc. announced that the company is working to bring the latest NVIDIA AI technology to VMware Cloud Foundation (VCF). Through this collaboration with NVIDIA, enterprises and cloud service providers will be empowered to build, deploy, and scale next-generation AI models on cutting-edge AI servers within their modern private cloud environments, all while leveraging the trusted operational experience of VCF.

Generative, agentic and physical AI applications are fundamentally transforming data center architectures, demanding unprecedented infrastructure capabilities. This expanded integration builds on the longstanding Broadcom and NVIDIA collaboration, adding support for NVIDIA Blackwell GPUs — including NVIDIA RTX PRO 6000 Server Edition GPUs — and NVIDIA networking. These broad options for NVIDIA accelerated computing address the need for increased choice, flexibility, and operational efficiency for AI infrastructure, enabling VCF customers to harness the full potential of NVIDIA’s AI innovation. These enhancements will expand the capabilities of the companies’ jointly developed VMware Private AI Foundation with NVIDIA.

"The emergence of generative AI is driving the need for a new level of accelerated computing infrastructure," said Justin Boitano, vice president of Enterprise AI at NVIDIA. "With NVIDIA networking technology and Blackwell GPUs, enterprises can build and deploy powerful AI applications directly within their existing private cloud, using the full capabilities of the NVIDIA platform in combination with VMware Cloud Foundation."

Integration of NVIDIA Blackwell Architecture GPUs
VCF will support NVIDIA’s latest GPU architecture, engineered for massive AI training, inference, and high-performance computing (HPC) at unprecedented scale. This includes specific support for:

High-Speed Networking with Enhanced DirectPath I/O
VCF will incorporate support for NVIDIA ConnectX-7 NICs and NVIDIA BlueField-3 400G DPUs with DirectPath I/O. This enables customers to leverage advanced capabilities like NVIDIA GPUDirect RDMA and NVIDIA GPUDirect Storage for high-speed, multi-node AI model training and data transfer, crucial for demanding Generative AI workloads.

Preservation of Core VCF Capabilities
A cornerstone of this integration is the ability for customers to deploy the latest NVIDIA innovations, including NVIDIA HGX servers, while retaining familiar VCF operational workflows and enterprise-grade virtualization features such as vMotion, High Availability (HA), Distributed Resource Scheduler (DRS), and Live Patching. This provides reliability and flexibility for mixed workloads, allowing AI/ML applications to run seamlessly alongside existing enterprise applications.

www.broadcom.com